Spatial Transcriptomics Deconvolution Cell2location

Visium and other spot-based spatial transcriptomics (Slide-seq, ST) cell type deconvolution with cell2location: load Space Ranger output, get a tissue-matched scRNA-seq reference from CELLxGENE, learn cell type signatures, estimate cell type abundance per spot with posterior uncertainty, and check the maps against marker genes. Default route for 'annotate cell types' on Visium when a GPU is available. Limits: needs a tissue- and state-matched reference; the default 30000 epochs take hours on CPU, so without a GPU it runs only with fewer epochs (exploratory maps) and can exceed one execution call's time limit. For fast per-spot proportions on CPU use spatial-mapping-tangram (clusters mode); for single-cell-resolution data (Xenium, MERFISH, CosMx) use spatial-transcriptomics-cell-type-annotation.

jaechang-hits Updated 389 repo stars
BixBench bioinformatics benchmark results — SciAgent-Skills achieves 92.0% accuracy

File contents

jaechang-hits/sciagent-skills commit 040d0c6a12

Frequently asked questions

npx get-skill add jaechang-hits/spatial-transcriptomics-deconvolution-cell2location